The existing test only verified the adaptive path. Add two more cases:
- enabled thinking (high): temperature must also be omitted
- no thinking (None): temperature must still be omitted
Made-with: Cursor
Two issues with DeepSeek V4 thinking mode support:
1. Missing thinking parameter injection.
DeepSeek V4 requires `extra_body: {"thinking": {"type": "enabled/disabled"}}`
— identical to VolcEngine/BytePlus. The code had this for volcengine,
byteplus, dashscope, minimax, and kimi but not DeepSeek. This means
`reasoning_effort=minimal` (thinking off) silently has no effect.
Root cause: the thinking-style→wire-format mapping was an if/elif chain
on provider *names*. DeepSeek was forgotten.
Fix: make the mapping declarative via `ProviderSpec.thinking_style`:
- "thinking_type" → {"thinking": {"type": "..."}} (DeepSeek, Volc, BytePlus)
- "enable_thinking" → {"enable_thinking": bool} (DashScope)
- "reasoning_split" → {"reasoning_split": bool} (MiniMax)
`_build_kwargs` now does a single dict lookup. Adding a new provider
with an existing wire format requires zero changes to the function.
2. Legacy session messages crash thinking-mode requests.
When a session was started without thinking mode (or with a different
model), assistant messages lack reasoning_content. DeepSeek V4 in
thinking mode rejects these with 400:
"The reasoning_content in the thinking mode must be passed back to the API."
This affects ALL assistant messages, not just those with tool_calls
(despite the docs only mentioning the tool_calls case).
Fix: `_build_kwargs` backfills `reasoning_content: ""` on every
assistant message missing it, but only when thinking mode is active.
This is semantically neutral — the model treats empty reasoning_content
as "no thinking happened on that turn". The backfill only touches the
in-memory request copy; session files on disk are untouched.
Tests: +5 (3 thinking toggle, 2 backfill). Full suite: 2377 passed.
Made-with: Cursor
#3412 stopped the headline raw_archive bloat but left four adjacent leaks
on the same pollution chain:
- archive() success path appended uncapped LLM summaries to history.jsonl,
so a misbehaving LLM could re-open the #3412 bug from the happy path.
- maybe_consolidate_by_tokens did not advance last_consolidated when
archive() fell back to raw_archive, causing duplicate [RAW] dumps of
the same chunk on every subsequent call.
- Dream's Phase 1/2 prompt injected MEMORY.md / SOUL.md / USER.md and
each history entry without caps, so any legacy oversized record (or an
unbounded user edit) would blow past the context window every dream.
- append_history itself had no default cap, leaving future new callers
one forgotten-cap-away from the same vector.
Changes:
- Cap LLM-produced summaries at 8K chars (_ARCHIVE_SUMMARY_MAX_CHARS)
before writing to history.jsonl.
- Advance session.last_consolidated after archive() regardless of whether
it summarized or raw-archived — both outcomes materialize the chunk;
still break the round loop on fallback so a degraded LLM isn't hammered.
- Truncate MEMORY.md / SOUL.md / USER.md and each history entry in Dream's
Phase 1 prompt preview (Phase 2 still reaches full files via read_file).
- Add _HISTORY_ENTRY_HARD_CAP (64K) as belt-and-suspenders default in
append_history with a once-per-store warning, so any new caller that
forgets its own tighter cap gets caught and observable.
Layer the caps by scope: raw_archive=16K, archive summary=8K,
append_history default=64K. Tight per-caller values cover expected
payloads; the wide default only catches regressions.
Tests: +9 regression tests covering each fix. Full suite: 2372 passed.
Made-with: Cursor
Cover two untested boundaries from #3412:
- _truncate_to_token_budget with positive budget exercises tiktoken
- _MAX_HISTORY_CHARS caps Recent History section in system prompt
Made-with: Cursor
Truncate the "Recent History" section injected by build_system_prompt()
to 32K chars. Without this, many accumulated history.jsonl entries could
still bloat the system prompt even with per-entry truncation in place.
Root cause: when consolidation LLM fails, raw_archive() dumped full message
content (~1MB) into history.jsonl with no size limit. Since build_system_prompt()
injects history.jsonl into every system prompt, all subsequent LLM calls exceeded
the 200K context window with error 1261.
Additionally, _cap_consolidation_boundary's 60-message cap caused consolidation
to get stuck on sessions with long tool chains (200+ iterations), triggering
the raw_archive fallback in the first place.
Three-layer fix:
- Remove _cap_consolidation_boundary: let pick_consolidation_boundary drive
chunk sizing based solely on token budget
- Truncate archive() input: use tiktoken to cap formatted text to the model's
input token budget before sending to consolidation LLM
- Truncate raw_archive() output: cap history.jsonl entries at 16K chars
Extend the existing on_progress callback to carry structured tool-event
payloads alongside the plain-text hint, so channels can render rich
tool execution state (start/finish/error, arguments, results, file
attachments) rather than only the pre-formatted hint string.
Changes
-------
- AgentLoop._tool_event_start_payload() — builds a version-1 start
payload from a ToolCallRequest
- AgentLoop._tool_event_result_extras() — extracts files/embeds from a
tool result dict
- AgentLoop._tool_event_finish_payloads() — maps tool_calls +
tool_results + tool_events from AgentHookContext into finish payloads
- _LoopHook.before_execute_tools() — passes tool_events=[...] to
on_progress together with the existing tool_hint flag
- _LoopHook.after_iteration() — emits a second on_progress call with
the finish payloads once tool results are available
- _bus_progress() — forwards tool_events as _tool_events in OutboundMessage
metadata so channel implementations can read them
- on_progress type widened to Callable[..., Awaitable[None]] on all
public entry points; _cli_progress updated to accept and ignore
tool_events
The contract is additive: callers that only accept (content, *, tool_hint)
continue to work unchanged. Callers that also accept tool_events receive
the structured data.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
``InlineKeyboardButton(label, callback_data=label)`` fails Telegram's
API when the label exceeds 64 bytes UTF-8. An LLM-generated long
option (realistic in multilingual flows) used to 400 the ``send_message``
call silently — user got nothing, agent heard a successful retry-then-drop.
Decouple display from wire: button text keeps the full label, callback_data
gets truncated at a UTF-8 char boundary. Tap echoes the prefix back as the
user message; the LLM understands a prefix of its own option just fine,
and the display the user saw was always the full string.
Locks: helper boundary behavior (ASCII, CJK, short labels pass through)
and end-to-end ``_build_keyboard`` integration with an over-cap label.
Made-with: Cursor
Buttons are semantic options, not a separate channel protocol: a user
who taps "Yes" and a user who types "yes" arrive at the agent as the
same string. Dropping ``msg.buttons`` when ``inline_keyboards=False``
was the worst of both worlds — the agent got told "Message sent with
N button(s)" while the user saw a question with no options.
Splice the labels into the message text instead. The LLM produces the
same ``message(buttons=...)`` call regardless of channel; the channel
layer picks the richest rendering it can afford — native keyboard when
enabled, bracketed inline text otherwise. Layout is preserved (one row
per line). Other channels can adopt the same helper incrementally.
Locks: canonical ``_buttons_as_text`` format, flag-off send-path
splices labels, flag-on send-path keeps content clean and rides
``reply_markup``.
Made-with: Cursor
Two kill-switch tests for the new inline-keyboards path. Neither is
flashy — they just make sure the next unrelated refactor can't quietly
regress two narrow contracts the PR relies on.
1. TelegramChannel._build_keyboard returns None whenever
TelegramConfig.inline_keyboards is False, even if buttons are
supplied. The flag defaults off; if someone ever flips that default
the change should fail this test before it reaches prod bots.
2. MessageTool rejects malformed `buttons` payloads (non-list, mixed
list/str row, non-str label, None label) up front instead of
letting them slip into the channel layer where Telegram would
silently 400 the send. Parametrized over four shapes the guard
needs to reject.
No production code touched.
Made-with: Cursor
Replace the dump→resolve→model_validate roundtrip with a recursive walk
that substitutes ${VAR} in string values directly on BaseModel /
__pydantic_extra__ / dict / list nodes. Identity is preserved on any
subtree with no references, so the original Config instance is returned
unchanged when nothing needs resolving.
Side effects:
- exclude=True fields (e.g. DreamConfig.cron) now survive even when
other fields in the same config contain ${VAR} references, closing
the edge case left open by the previous fast-path-only fix.
- _has_env_refs is dropped (the walker short-circuits naturally).
- Added a regression test pairing cron with a resolved providers.groq
api_key to lock the coexistence case.
Made-with: Cursor
`resolve_config_env_vars` unconditionally dumped the config via
`model_dump(mode="json")` and revalidated it, which silently dropped
any field declared with `exclude=True` (e.g. `DreamConfig.cron` —
introduced by the Dream rename refactor in #2717). Result:
`agents.defaults.dream.cron` was never honored at runtime — the gateway
always fell back to the default `every 2h` schedule even when `cron`
was set in config.json.
Fix: skip the roundtrip entirely when the config has no `${VAR}`
references. Env-var interpolation still works unchanged when refs
exist; the legacy `cron` override now survives the common case of
fully-resolved config.
Regression test covers the bug path.
Adds a focused regression test so the fix for tool_result image
handling cannot silently revert. Two cases:
- list content with an image_url + text block -> image_url is
translated to a native Anthropic image block, sibling text passes
through unchanged
- plain string content passes through untouched (the new list branch
must not alter the string path)
These cover the exact symptom surface (silent image drop with a
"Non-transient LLM error with image content" warning) and the only
two content shapes tool results actually take today.
Made-with: Cursor
_tool_result_block passed list content through unchanged, so image_url
blocks returned by tools (e.g. read_file on an image file, which
returns OpenAI-format image_url blocks via build_image_content_blocks)
reached the Anthropic API unconverted and were rejected. User-role
messages already ran through _convert_user_content at the call site,
so inbound Telegram photos worked, but tool results did not.
Run _convert_user_content on list content inside _tool_result_block
so image_url blocks become native Anthropic image blocks. Required
making _convert_user_content a @staticmethod (it did not use self)
and calling _convert_image_block via the class to match.
Repro: an agent calling read_file on any image file got a
"Non-transient LLM error with image content, retrying without images"
warning and the image was silently dropped from the conversation.
When the main agent spawns multiple sub-agents, each completion
independently triggered a new _dispatch, causing 3-4 user-visible
responses instead of a single comprehensive report.
- Extend _drain_pending to block-wait on pending_queue when sub-agents
are still running, keeping the runner loop alive for in-order injection
- Pass pending_queue in the system message path so subsequent sub-agent
results can still be injected mid-turn via a new dispatch
Locks in the four behaviors introduced by the fix so they can't silently
revert:
- _should_use_responses_api accepts github_copilot on its non-OpenAI base
- _build_responses_body strips the 'github_copilot/' routing prefix
- /responses failures on github_copilot do not fall back to /chat/completions
Made-with: Cursor
Calling GitHub Copilot with `gpt-5.*` / `o*` models (e.g.
`github_copilot/gpt-5.4`, `github_copilot/gpt-5.4-mini`) failed with a
chain of misleading errors:
1. `Unsupported parameter: 'max_tokens' is not supported with this
model. Use 'max_completion_tokens' instead.`
2. `model "gpt-5.4-mini" is not accessible via the /chat/completions
endpoint` (`unsupported_api_for_model`).
3. `The requested model is not supported.` (`model_not_supported`)
even after routing to /responses.
Root causes (each one masked the next):
* The `github_copilot` ProviderSpec did not opt into
`supports_max_completion_tokens`, so `_build_kwargs` always sent the
legacy `max_tokens` parameter that GPT-5/o-series reject.
* `_should_use_responses_api` was hard-gated to
`spec.name == "openai"` plus a direct-OpenAI base URL, so the
GitHub Copilot backend always went through /chat/completions even
for models the Copilot gateway exposes only via /responses
(e.g. `gpt-5.4-mini`).
* When /responses did fail on github_copilot, the existing
"compatibility marker" heuristic silently fell back to
/chat/completions — which can never succeed for these models — so
the real upstream error was hidden.
* `_build_responses_body` did not honour `spec.strip_model_prefix`,
so the request body sent `model="github_copilot/gpt-5.4-mini"`
(with the routing prefix), which the Copilot gateway rejects with
`model_not_supported`. (`_build_kwargs` already stripped it; this
branch was missed.)
Fix:
* registry.py: set `supports_max_completion_tokens=True` on the
`github_copilot` spec so requests use `max_completion_tokens`.
* openai_compat_provider.py:
- `_should_use_responses_api` now also allows the
`github_copilot` spec, and skips the direct-OpenAI base check
for it (the Copilot gateway is its own base URL).
- `_build_responses_body` now strips the model routing prefix
when `spec.strip_model_prefix` is set, matching `_build_kwargs`.
- `chat` / `chat_stream` no longer fall back from /responses to
/chat/completions on the `github_copilot` spec: the fallback
cannot succeed for GPT-5/o-series and would mask the real
gateway error.
Tests:
* tests/cli/test_commands.py: switched the
`test_github_copilot_provider_refreshes_client_api_key_before_chat`
fixture model from `gpt-5.1` to `gpt-4` so it continues to exercise
the /chat/completions code path it was designed for (gpt-5.1 now
correctly routes to /responses on github_copilot).
* `pytest tests/providers/ tests/cli/test_commands.py` — 314 passed.
* Verified end-to-end against the live Copilot gateway with both
`github_copilot/gpt-5.4` and `github_copilot/gpt-5.4-mini`.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
On Windows, opening a directory with O_RDONLY raises PermissionError.
Wrap the directory fsync in a try/except PermissionError — NTFS journals
metadata synchronously so the directory sync is unnecessary there.
Also adjust test assertions to expect 1 fsync call (file only) on
Windows vs 2 (file + directory) on POSIX.
On filesystems with write-back caching (rclone VFS, NFS, FUSE mounts)
the OS page cache may buffer recent session writes. If the process is
killed before the cache flushes, the most recent conversation turns are
silently lost — causing the agent to "forget" recent context and
respond to stale history on the next startup.
Changes:
- session/manager.py: add fsync=True option to save() that flushes the
file and its parent directory to durable storage. Add flush_all() that
re-saves every cached session with fsync. Default save() behavior is
unchanged (no fsync) to avoid performance regression in normal
operation.
- cli/commands.py: call agent.sessions.flush_all() in the gateway
shutdown finally block, after stopping heartbeat/cron/channels.
- tests/session/test_session_fsync.py: 8 tests covering fsync flag
behavior, flush_all with empty/multiple/errored sessions, and
data survival across simulated process restart.
- tests/cli/test_commands.py: add sessions attribute to _FakeAgentLoop
so the gateway health endpoint test passes with the new shutdown
flush.
DashScope rejects the OpenAI-style value "minimal" with
`'reasoning_effort.effort' must be one of: 'none', 'minimum', 'low',
'medium', 'high', 'xhigh'`, but nanobot was passing the string through
verbatim. Users who tried the documented "minimal" to disable thinking
got a 400; users who tried the DashScope-native "minimum" to work
around it got `enable_thinking=True` because the internal comparison
was a hard string match on "minimal".
Introduce a semantic/wire split in `_build_kwargs`:
- `semantic_effort` is the internal canonical form (OpenAI vocabulary).
"minimum" on the way in is normalized to "minimal" here so both
spellings share one meaning.
- `wire_effort` is what we actually serialize. For DashScope with
semantic_effort == "minimal" we translate to "minimum" on the way
out; other providers are unchanged.
- `thinking_enabled` and the Kimi thinking branch now compare on
`semantic_effort`, so either user spelling correctly disables
provider-side thinking.
Tests:
- Strengthen `test_dashscope_thinking_disabled_for_minimal` to assert
the wire value is "minimum" in addition to the extra_body signal;
the original version only checked extra_body and let the
invalid-value bug slip through.
- Add `test_dashscope_thinking_disabled_for_minimum_alias` so a user
who read the DashScope docs and configured "minimum" still gets
thinking off.
- Add `test_non_dashscope_minimal_not_retranslated` to pin down that
the DashScope-specific translation does not leak to OpenAI et al.
- Add ISO-639 pattern validation (2-3 lowercase letters) to schema
- Normalize empty language to None in provider constructors
- Extract shared httpx mock stubs, parameterize provider tests
- Add test for language=None omitting field from multipart body
- Add test for Pydantic pattern validation rejecting invalid codes
Wire up the existing office document extractors in document.py to
ReadFileTool by adding an extension guard and _read_office_doc() method
that follows the established PDF pattern. Handles missing libraries,
corrupt files, empty documents, and 128K truncation consistently.
Non-priority slash commands (e.g. /new, /help, /dream-log) arriving
while a session has an active LLM turn were silently queued into the
pending injection buffer and later injected as raw user messages into
the LLM conversation. This caused the model to respond to "/new" as
plain text instead of executing the command.
Root cause: the run() loop only checked priority commands (/stop,
/restart, /status) before routing messages to the pending queue. All
other command tiers (exact, prefix) bypassed command dispatch entirely.
Changes:
- Add CommandRouter.is_dispatchable_command() to match exact/prefix
tiers, mirroring the existing is_priority() pattern.
- In run(), intercept dispatchable commands before pending queue
insertion and dispatch them directly via _dispatch_command_inline().
- Extract _cancel_active_tasks() from cmd_stop for reuse; cmd_new now
cancels active tasks before clearing the session to prevent shared
mutable state corruption from concurrent asyncio coroutines.
- Update /new semantics: stops active task first, then clears session.
- Update documentation in help text, docs, and Discord command list.
Problem:
Modern LLMs (GPT-5.4, Claude, Gemini) produce markdown-heavy responses with
numbered lists, headers, and nested formatting. The Telegram channel's
_markdown_to_telegram_html() converter has gaps that leave these poorly
formatted:
1. Numbered lists (1. 2. 3.) have zero handling — sent as raw text
2. Headers (# Title) are stripped to plain text, losing visual hierarchy
3. Mid-stream edits send raw markdown (users see **bold** and ### headers
while the response generates, before the final HTML conversion)
Root Cause:
_markdown_to_telegram_html() handles bullets (- *) but skips numbered lists
entirely. Headers are stripped of # but not given any emphasis. The streaming
path in send_delta() sends buf.text as-is during mid-stream edits (plain
text, no parse_mode) — only the final _stream_end edit converts to HTML.
Fix:
1. Headers now render as <b>bold</b> in the final HTML (using placeholder
markers that survive HTML escaping, restored after all other processing)
2. Numbered lists are normalized (extra whitespace after the dot is cleaned)
3. New _strip_md_block() function strips markdown syntax for readable
plain-text preview during streaming mid-edits
The final _stream_end HTML conversion is unchanged — it still produces
full HTML with parse_mode=HTML. Only the intermediate edits are improved.
Tests:
Added 10 new tests covering:
- Headers converting to bold HTML
- Numbered list preservation and whitespace normalization
- Headers with HTML special characters
- Mixed formatting (headers + bullets + numbers + bold)
- _strip_md_block for inline formatting, headers, bullets, numbers, links
- Streaming mid-edit markdown stripping (initial send + edit)
ZhiPu API returns code 1302 with Chinese text "速率限制" instead of
standard HTTP 429 + "rate limit", causing the retry engine to treat
it as non-transient and fail immediately.
